In a recent opinion presented to the Court of Justice of the European Union, First Advocate General Maciej Szpunar warned that Google’s AI chatbot, Gemini, could breach press publishers’ copyright rights if it retrieves and reproduces online news articles in a way that makes the original publications unnecessary. The opinion follows a complaint lodged by the Hungarian media company Like Company against Google Ireland over alleged use of its content. The case centers on whether Gemini’s handling of news material violates European Union law.
Gemini functions as a large language model (LLM), a system that learns from vast amounts of text data to predict and generate human‑like responses. By training on diverse content, Gemini can generate answers to user queries that incorporate information from the texts it has processed.
Like Company, which operates several news portals, alleged that Google’s chatbot accessed its articles, used them to train Gemini’s models, and reproduced their text for users. The Hungarian regional court referred the matter to the European Court, seeking guidance on the legality of using press articles for AI training and answer generation.
Legal Assessment of AI Content Use
First Advocate General Szpunar’s opinion explains that potential infringement hinges on two key factors: unlawful acquisition of the article by bypassing technical safeguards such as paywalls, and the extent to which the chatbot reproduces the original expression rather than merely summarizing facts.
The adviser notes that if Gemini retrieves an article without respecting the authorisation mechanisms in place, such as artificially ignoring or circumventing paywalls, the AI system may be engaging in unlawful use of the content.
In addition, the opinion stresses that reproducing an article’s distinctive expression in a recognizable form can amount to infringement, especially when the reproduction makes it unnecessary for users to consult the original publication to obtain that information.
When the chatbot systematically incorporates the same expression across related user queries, the adviser argues that publishers’ rights to make their content available to the public can be infringed, as the model’s answers effectively replace the need to subscribe or visit the publisher’s site.
If such practices are embedded in the business model of the chatbot provider, or if the provider fails to implement adequate safeguards to prevent repeated reproduction of protected content, the provider could face liability under EU copyright law.
The opinion therefore calls for a clearer definition of permissible AI training and content use within the European framework, prompting a reevaluation of existing copyright regimes in the context of rapidly evolving AI technologies.
For news publishers, the adviser’s warning highlights the growing risk that AI‑generated content could undermine subscription models and the revenue streams that fund investigative journalism.
For AI developers, the legal assessment suggests the need to adopt robust content filtering and licensing mechanisms to mitigate potential infringement claims.
As the Court of Justice considers the matter, the outcome will set a precedent for how AI systems interact with copyrighted news material across the EU, influencing both industry practices and legislative initiatives.
While Google has not yet issued a statement in response to the opinion, stakeholders in both the journalism and technology sectors will closely monitor the court’s ruling for its implications on content licensing and AI ethics.
In the broader context of AI development, the case underscores the importance of aligning advanced computational tools with existing intellectual property frameworks to ensure fair use while protecting legitimate copyright interests.
